Abstract
Objective. We aimed to evaluate how minimal (clinically) important differences (MCID/MID) were calculated in rheumatology in the past 2 decades and demonstrate how the calculation is compromised by the lack of interval scaling.
Methods. We conducted a systematic literature review on articles reporting MCID calculation in osteoarthritis (OA) and rheumatoid arthritis (RA) from January 1, 1989, to May 9, 2014. We evaluated the methods of MCID calculation and recorded the ranges of MCID for common patient-reported outcome measures (PROM). Taking data from the Health Assessment Questionnaire (HAQ), we showed the effects of performing mathematical calculations on ordinal data.
Results. A total of 330 abstracts were reviewed and 123 articles chosen for full text review. Thirty-six (19 OA, 16 RA and 1 OA-RA) articles were included in the final evaluation. The Western Ontario and McMaster Universities Osteoarthritis Index (WOMAC) was the most frequently reported PROM with relevant calculations in OA, and the HAQ in RA. Sixteen articles used anchor-based methods alone for calculation of MCID, and 1 article used distribution-based methods alone. Nineteen articles used both anchor and distribution-based methods. Only 1 article calculated MCID using an interval scale. Wide ranges in MCID for the WOMAC in OA and HAQ in RA were noted. Ordinal-based derivations of MCID are shown to understate true change at the margins, and overstate change in the mid-range of a scale.
Conclusion. The anchor-based method is commonly used in the calculation of MCID. However, the lack of interval scaling is shown to compromise validity of MCID calculation.
